Learning a representation of a believable virtual character's environment with an imitation algorithm.

Fabien Tencé (1) , Cédric Buche (2) , Olivier Marc Marc (1) , Pierre de Loor (3, 4)

1 LISYC - Laboratoire d'Informatique des Systèmes Complexes
2 Lab-STICC_ENIB_CID_IHSEV
3 Lab-STICC - Laboratoire des sciences et techniques de l'information, de la communication et de la connaissance
4 CERV - Centre Européen de Réalité Virtuelle

Improving the believability of the characters populat- ing a virtual environment is one way to make users of feel in the simulation. The model developed in (Le Hy et al. 2004) seems to be a good base to generate be- lievable behaviours: it is close to what gives quite good results in the industry but also uses probabilities and a learning algorithm which could make the illusion of believability last longer. This article rst describes how this model works. As it uses a manually-de ned repre- sentation of the character's environment, improvements can be done to that representation to enhance its auton- omy and believability. We propose to use an other model named growing neural gas to learn by imitation the rep- resentation of the environment. The way this model was implemented and the evaluation of the quality of the learned representations are then detailed. Ideas about improvements for the growing neural gas to give more information to Le Hy's model are given in the conclu- sion.

Domaines: Intelligence artificielle [cs.AI]

Fabien Tencé, Cédric Buche, Olivier Marc Marc, Pierre de Loor. Learning a representation of a believable virtual character's environment with an imitation algorithm.. First annual Pan-Arabic International Conference on Intelligent Games and Simulation (GAMEON-ARABIA'10), Dec 2009, Egypt. pp.141-145. ⟨hal-00784296⟩